A risk-taking philosophy for entering a new, unproven product category: instead of trying to prevent all failure before launch, the organization accepts imperfection at launch and invests its effort in detecting and recovering from failure quickly.
Elizabeth Stone illustrates this with Netflix's rollout into live streaming — the org didn't wait for a fully polished live-streaming capability before shipping; it launched, expected rough edges, and built the muscle to react and fix fast rather than over-engineer upfront certainty.
This complements Blameless Retro (learning from failure without blame) and depends on the same trust fabric as Highly Aligned, Loosely Coupled and Non-Interference on Non-Material Decisions — fast recovery requires that teams are empowered to act and correct without waiting for top-down sign-off.
Applied to agent sessions specifically: when an agent goes off track or gets uncooperative, the default move is to close the session outright rather than argue with it or steer it harder. "Close the session. And then I improve the evals and then I try from scratch again." The improvement happens in the evals/success criteria, not in the abandoned session, so the next attempt starts from a better specification rather than patched-up context.